A Six Degree of Freedom Micromanipulator for Ophthalmic Surgery

Kenneth W.Grace, Mechanical Engineering

J. Edward Colgate, Mechanical Engineering
Matthew R. Glucksberg, Biomedical Engineering
John H. Chun, Mechanical Engineering
Northwestern University Evanston, IL 60208
not useful for guiding the hypodermic needle in the way
described, since they exhibit Cartesian movements. The
spherical movement needed can be found in some
commercially available micromanipulators, and some
have been constructed for this type of research [9], but
they are circular track serial devices, (figure I), which are
mechanically constrained to move on a spherical surface.

Abstract
Needs which have arisen in ophthalmic research have
motivated the development of a six degree of freedom
parallel micromanipulator. The first application of the
tool will be in the treatment of retinal vemus occlusion,
for which micron-scale spherical movement of a glass
micropipette tip inside the eye is required. The initial
speration mode will be open loop while future operation
will be in a force-reflecting bilateral (macro-master /
micro-slave) arrangement. Presented here are some of
the design criteria, the mathematical tools used in
evaluating various parallel manipulator designs, and the
final kinematic configuration.

1. Background and Motivation

An active area of research is the treatment of retinal
venous occlusion by delivery of anticoagulant drugs
directly to blood clots in retinal vessels [1,2,41.
Shortcomings of micro manipulators currently being used
in such research have motivated the development of a new
tool. In order to better explain the design constraints,
some steps which are included in a typical procedure are
outlined here. A hypodermic needle is inserted through
the wall of the eye to gain access to the interior surface.
A rigid glass micropipette is inserted through the
hypodermic needle far enough that its tip protrudes
beyond the end of the needle. The micropipette is guided
to various injection sites on the retina by pivoting the
hypodermic needle about its puncture point, that is, the
point of intersection between the needle and the wall of
the eye. By keeping the needle nearly stationary at the
point where it passes into the eye's interior, damage to the
wall of the eye is minimized. A third allowable degree of
freedom is movement along the axis of the hypodermic
needle since it does not deform the wall of the eye at the
puncture point significantly.
Most commercially available micromanipulators are

:igure 1: Schematic of a traditional sphericalmovement micromanipulator.

630

1050-4729/93$3.00 0 1993 IEEE

summarized several advantages not listed here [7].)

Different geometries were evaluated through the
definition of six quantities derived from the jacobian.
Details follow.

Such micromanipulators tend to be bulky because the

mechanical tracks span the entire range of motion at all
times, prohibiting the accommodation of multiple
simultaneous entries into the eye. Another point of
functional inflexibility is that the center of spherical
motion can not be moved with respect to the base of these
manipulators due to the fixed radii of the mechanical
tracks.
As an alternative to these gimbal-type devices which
are physically constrained to the desired "puncturecentered movement, the six DOF parallel device can be
constrained mathematically by its computer controller.
The advantage is not only the versatility gained by
redundant degrees of freedom, but also that of
compactness. A desirable size and shape for the device is
that of a human wrist and hand gripping a slender tool.
This will allow multiple manipulators to be used
simultaneously on a single eye when necessary.
The initial operation mode will be open loop wherein
the operator will watch the end effector (through the
pupil with the aid of a microscope) and guide it using a
multi-dimensional joystick input device connected to a
computer controller. Subsequent modes of operation
will include force-reflecting bilateral control
arrangement (discussed later in this paper).

2.2. Coordinate Systems

The traditional scheme of specifying end effector

orientation with three euler angles will be slightly
modified to better examine each manipulator design's
suitability for the task at hand. Imagine a third
coordinate system [U*V*W*] in addition to those
pictured in figure 2 whose origin is always coincident
with the [VVW] origin at the puncture point, and whose
U*V* plane is always parallel to the XY plane. As the
hypodermic end effector with rigidly attached system
[UVW] is placed in some arbitrary orientation, the
position of the tip of the needle can be projected onto the
U*V* plane. (Recall that the puncture point is not, in
general, at the tip of the needle.) The two coordinates
(u*,v*) necessary to describe this projected tip position
make up two of the three orientation variables. The third
will be the angle of rotation y of the platform about the
hypodermic. The position (x.y,z) of the puncture point in
the [XYZ] frame completes the set of six variables needed
to fully specify the position and orientation of the end
effector.
These choices were shaped by the microinjection task.
During a procedure, the greatest part of the movement
will be in the (u*,v*) coordinates, while rotation y about
the needle has little effect and therefore is of little
importance for this application.

2. Kinematic Design
2.1. Initial Choice

The Stewart Platform [lo] parallel six DOF robot

scheme (figure 2) was chosen as a starting point for its
inherent stiffness and compactness. (Merlet has

Platform
\

2.3. Sensitivity Parameters

One important requirement in choosing a specific

kinematic design was low sensitivity of endpoint position
to perturbations in leg length, because amplification of
uncertainty in joint position is undesirable from the
precision standpoint. Since the jacobian maps joint
perturbations into endpoint perturbations, each Jacobian
element can be thought of as a gain from joint to endpoint
space.

Legs are Prismatic Joints

U l (Ball-Jointed at Each End)

[i]
6V*

cy

Endpoint
Qlange

Jacobian

Joint
-ge

The sum of the absolute values of all Jacobian elements

in a single row, (or the one-norm of each row vector) is
some measure of a "worst-case" gain, mapping a

'igure 2: Stewart platform schematic

631

simultaneous unit perturbation in all joints to a

perturbation in one of the six endpoint coordinates. For
the fnst row of the jacobian, define the sensitivity Sl to
be

Sensitivity of
X

2
2
4
2

=kl&l

s1 i=1

The vector comprising all six such sensitivity measures

will be called the sensitivity vector for convenience in
this paper. Since the sensitivity vector is different for
each position and orientation of the end effector of a given
design, a few key positions were chosen as representative
and considered across all designs. It should also be noted
that comparison of sensitivity vectors is not a purely
quantitative task, since one geometry might yield high
sensitivity in one degree of freedom and low in another
while the converse might be true in a second geometry.
Knowledge of the movements that would be most critical
in the microinjection application shaped the judgments of
which sensitivity measures to weigh more heavily, (such
as (u*,v*)), and which to, in some cases, ignore completely
(such as y). Yet another important point to note is that,
while the vectors can be compared to one another,
elements within one vector can not be directly compared.
This is because the elements of the sensitivity vector
corresponding to the rotational degree of freedom has
units of (radians/length) while the translational DOF
units are (lengwength).

2
4
2
2

2
2
2
6

5.94 5.21
7.48 7.67
3.42 3.81
12.11 12.43

z
1.07
1.24
1.28
1.01

v*

0.71
0.86
0.86
0.67

0.77
0.76
0.80
0.72

y
3.01
1.74
1.61
8.48
I

Table 1: Effect of three geometry parameters

on the sensitivity vector.

The U*and v* values, which are of primary importance for

microinjection, are best when the manipulator is slender
(row 4). The tradeoff is a significant degradation in x and
y (using the first row as a reference). If the platform is
large compared to the base (row 2), a similar but less
severe tradeoff results. The third row shows a conical
design in which the base is twice the diameter of the
platform. The x and y sensitivities are better than in the
case of the cylindrical design of row 1, and U*and v* are
not significantly worse. This design was rejected,
however, in favor of row 1 on the basis of range of motion
(which always competes with low sensitivity).

3. Physical Design
3.1. Stewart Platform

2.4. Mathematical Design Search Method

Approaches to the physical design of a Stewart Platform

can be divided into two classes: (1) legs either pull or
push against an antagonistic force (i.e. a spring pushes the
platform away from the base at all times while six
variable length cables constrain it [51), or (2) each leg is a
linear actuator which can be commanded to a desired
length.
In order to get a reasonable range of motion and good
response characteristics using approach (1) above, a
considerable amount of potential energy must be stored in
the antagonistic system. Failure of a leg connection in
such a system could have catastrophic results in
ophthalmic surgery. With approach (2), the biggest
problems are the actuator's overall length, length to
stroke ratio, and diameter to length ratio. Most of these
quantities are much larger than desired with
commercially available linear actuators, leading to
problems with leg collisions, high sensitivity values, and
poor range of motion.

The approach, using the symbolic manipulation

software package, Mathematica, was to leave the Stewart
Platform geometry parameters of interest and the set of
endpoint coordinates as variables, and then to solve the
inverse kinematics problem of finding the joint positions
(leg lengths) as a function of those variables. This is
straightforward for a parallel manipulator, because once
the coordinates of all leg attachment points are known in
a single coordinate frame, leg lengths are found by vector
differences. The inverse jacobian is then obtained by
symbolically differentiating with respect to the joint
positions. From this state, endpoint position and
geometry values can be supplied, and the inverse jacobian
evaluated and inverted to give the jacobian.
Following this procedure, many geometries and
endpoint positions can be examined without having to
invert or differentiate with numerical methods.
Parameters to be chosen included the diameter, Db, of the
smallest circle circumscribing the leg attachments to
ground (base size), a similar circle diameter, Dp, for the
upper end leg attachments (platform size), and a nominal
value, H, for the distance between platform and base.
Two general areas of investigation into geometry
effects were the ratio H:Dp (slenderness), and the ratio
Dp:Db (conicalness). Table 1 shows a few examples of
the effects of geometry on sensitivity vectors.

3.2. Variations on Initial Design

To circumvent some of these difficulties, a variation

was devised (figure 3). This variation allows the linear
actuators to remain fixed to ground rather than being ball
jointed to ground, which in turn generates fewer
prbblems with leg collisions and sensitivity parameter

632

goals. This same variation was presented by Merlet [6]

with the minor difference that Merlet's platform-end leg
connections occur in three coincident pairs, while figure 3
shows six distinct platform attachment points.
It was also decided that the needle should not be
mounted perpendicular to the platform in order to avoid
collisions between the back end of the micropipette and
the legs. (Recall that the micropipette is rigid and must
pass all the way through the hypodermic needle.)
Mounting at 45 degrees not only avoids collisions but
makes micropipette access much more convenient yielding
a more functional design.

4. Future Work
Following successful open loop operation, the
micromanipulator will be fitted with end effector force
sensing hardware and connected to a force reflecting
macro manipulandum (already developed)[81. From this
macro/mim bilateral manipulator platform, research in
impedance shaping bilateral control can proceed. The goal
of this research, the theory for which is described in detail
in other papers [3], is to develop a computer mediated hand
tool which reflects a "realistic feel" of the environment
being encountered by the micromanipulator while
performing a task on a level below that of human
dexterity.

5. Acknowledgments
The authors gratefully acknowledge the support of The
Margaret W. and Herbert Hoover Jr. Foundation in
funding this project.

i w

%ure 3:

Schematic of the Merlet Parallel

Manipulator with 45 degree hypodermic needle

end effector.
These variations on the earlier manipulator design
were not without cost. The sensitivity vector suffered
somewhat. After considering the sensitivities, range of
motion and function, the final design now has leg length,
platform diameter, and base diameter all of two inches
yielding a sensitivity vector of (5.31,5.41, 1.50,0.48,0.89,
1.87). The physical design is also complete, and the
manipulator is being fabricated (figure 4). It will be
actuated by brush-type D.C. torque motors and ball
screws.

633

c)

.-n
El

.-U
E

E0

I 2
c,

U
X Q

.-cn
Q

)
I

c
cn

a,

0
Q)

Q)

c)

U)
CI

.-E

I&

U
Q)
L.

39
m0

Gc
634

References
[l] B.E. AUf. and E. De Juan Jr. In Vivo Cannulationof
Retinal Vessels. Graefe's Arch Clin Exp Ophthalmol.
225221-225.1987.
[2] B.Becker and L. P. Post. Retinal Vein Occlusion.
American Journal of Ophthalmology,34677-6861951.
[3] J. E. Colgate. Power and Impedance Scaling in
Bilateral Manipulation. Proc. 1991 IEEE Int'l Con$ on
Robotics and Automation 3:2292-2297.
[4] M. R. Glucksberg. and R. Dum. Direct Measurement
of Retinal Microvascular Pressure in the Live
Anesthetized Cat. Microvascular Research. in press,
1992.
[5] S. E. Landsberger and T. B. Sheridan. A Minimal,
Minimal Linkage: The Tension-CompressionParallel
Link Manipulator. Proceedings of the IMACS/SICE
International. Symposium on Robotics, Mechatronics and
Manufacturing System '92 Kobe, 1:493-500,1992.

[6] J-P. Merlet. Direct Kinematics and Assembly Modes

of Parallel Manipulators. Int'l Journal of Robotics
Research, 11 (2):150-162, 1992.
[7] J-P. Merlet. Parallel Manipulators: State of the Art
and Perspectives. Proceedings of the IMACYSICE
International. Symposium on Robotics, Mechatronics and
Manufacturing Systems '92 Kobe, 1:403-408, 1992.
[8] P. A. Millman and J. E. Colgate. Design of a Four
Degree-of-FreedomForce-Reflecting Manipulandum
with a Specified ForcdTorque Workspace, Proc. I991
IEEE Int'l Conf: on Robotics and Automation, 2 14881493.
[9] C. J. Pournaras, R. D. Shonat, J-L. Munoz, and B.L.
Petrig, New Ocular Micromanipulator for
Measurements of Retinal and Vitreous Physiologic
Parameters in the Mammalian Eye. Experimental Eye
Research, 53:723-727, 1991.
[lo] D. Stewart. A Platform with 6 Degrees of Freedom.
Proc. Inst. Mech. Engr., 180371-386, 1965.

635